Quantization made by Richard Erkhov.

Qwen2-0.5B-DPO - GGUF

Model creator: https://huggingface.co/trl-lib/
Original model: https://huggingface.co/trl-lib/Qwen2-0.5B-DPO/

Name	Quant method	Size
Qwen2-0.5B-DPO.Q2_K.gguf	Q2_K	0.32GB
Qwen2-0.5B-DPO.IQ3_XS.gguf	IQ3_XS	0.32GB
Qwen2-0.5B-DPO.IQ3_S.gguf	IQ3_S	0.32GB
Qwen2-0.5B-DPO.Q3_K_S.gguf	Q3_K_S	0.32GB
Qwen2-0.5B-DPO.IQ3_M.gguf	IQ3_M	0.32GB
Qwen2-0.5B-DPO.Q3_K.gguf	Q3_K	0.33GB
Qwen2-0.5B-DPO.Q3_K_M.gguf	Q3_K_M	0.33GB
Qwen2-0.5B-DPO.Q3_K_L.gguf	Q3_K_L	0.34GB
Qwen2-0.5B-DPO.IQ4_XS.gguf	IQ4_XS	0.33GB
Qwen2-0.5B-DPO.Q4_0.gguf	Q4_0	0.33GB
Qwen2-0.5B-DPO.IQ4_NL.gguf	IQ4_NL	0.33GB
Qwen2-0.5B-DPO.Q4_K_S.gguf	Q4_K_S	0.36GB
Qwen2-0.5B-DPO.Q4_K.gguf	Q4_K	0.37GB
Qwen2-0.5B-DPO.Q4_K_M.gguf	Q4_K_M	0.37GB
Qwen2-0.5B-DPO.Q4_1.gguf	Q4_1	0.35GB
Qwen2-0.5B-DPO.Q5_0.gguf	Q5_0	0.37GB
Qwen2-0.5B-DPO.Q5_K_S.gguf	Q5_K_S	0.38GB
Qwen2-0.5B-DPO.Q5_K.gguf	Q5_K	0.39GB
Qwen2-0.5B-DPO.Q5_K_M.gguf	Q5_K_M	0.39GB
Qwen2-0.5B-DPO.Q5_1.gguf	Q5_1	0.39GB
Qwen2-0.5B-DPO.Q6_K.gguf	Q6_K	0.47GB
Qwen2-0.5B-DPO.Q8_0.gguf	Q8_0	0.49GB

Original model description:

base_model: Qwen/Qwen2-0.5B-Instruct datasets: trl-lib/Capybara-Preferences library_name: transformers model_name: dpo-qwen2 tags: - generated_from_trainer - trl - dpo licence: license

Model Card for dpo-qwen2

This model is a fine-tuned version of Qwen/Qwen2-0.5B-Instruct on the trl-lib/Capybara-Preferences dataset. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="qgallouedec/dpo-qwen2", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with DPO, a method introduced in Direct Preference Optimization: Your Language Model is Secretly a Reward Model.

Framework versions

TRL: 0.12.0.dev0
Transformers: 4.45.0.dev0
Pytorch: 2.4.1
Datasets: 3.0.0
Tokenizers: 0.19.1

Citations

Cite DPO as:

@inproceedings{rafailov2023direct,
    title        = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
    author       = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
    year         = 2023,
    booktitle    = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
    url          = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
    editor       = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
}

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}